Anthropic Chat Model node — Add cache_control / prompt caching support

Feature Requests
1

Anthropic Chat Model node — How to enable prompt caching (cache_control)?

The idea is:

My use case:

I’m running n8n Cloud (v2.19.3) in production with @n8n/n8n-nodes-langchain.lmChatAnthropic
(typeVersion 1.5) using Claude Sonnet 4.6. My setup is an AI sales assistant for a car
dealership in Brazil that handles WhatsApp customer conversations 24/7.
The system prompt is ~5,600 tokens and is sent to Anthropic on every customer message.
With prompt caching enabled (cache_control: { type: "ephemeral" }), I would save
~50-70% on input token costs per call.
Currently, the node does NOT expose cache_control — only temperature, topK, topP,
maxTokensToSample, and thinkingMode are available. The underlying @langchain/anthropic
library DOES support cache_control natively, but the n8n wrapper doesn’t expose it.
I confirmed via email with n8n support that there’s no documented or undocumented
workaround for the current node.

I think it would be beneficial to add this because:

  1. Direct cost reduction (~50-70% on Anthropic input tokens) for any user with
    system prompts >1024 tokens — common for AI agent use cases.
  2. The underlying library (@langchain/anthropic) already supports cache_control,
    so implementation on n8n’s side is small (a few lines in message construction).
  3. Anthropic prompt caching has been available since 2024 and is now a standard
    feature most production users expect.
  4. n8n Cloud users running AI agents with rich system prompts (rules, examples,
    tool descriptions) have no path to reduce these costs without rebuilding agent
    logic via raw HTTP Request Node — losing memory/tool orchestration.
  5. Other LLM nodes in the n8n ecosystem could benefit from similar caching
    support over time (e.g., OpenAI prompt caching).

Any resources to support this?

Are you willing to work on this?

Not in a position to contribute code (not a TypeScript developer), but happy to:

  • Test the feature in production once available
  • Provide before/after cost data for the n8n team
  • Write user documentation / blog post about the cost savings